OpenAI and Anthropic prepare for anti-AI unrest scenarios

The scenarios on the table

The exercises resemble military planning sessions: employees work through possible events and how the companies might respond. The scenarios reported by Axios include:

OpenAI said it runs exercises covering a range of possible circumstances. A company representative said the scenarios are not considered inevitable, but are a way to prepare for different situations.

Anthropic’s earlier response

Anthropic declined to comment on the exercises. But the prospect of unrest fits with a step the company was reported to be taking in September: building a system to monitor anti-AI protest movements and gather intelligence, using an approach described as “preventing crimes before they happen.”

The article also points to an earlier, less successful response by tech billionaires, including Mark Zuckerberg, who staged public-facing efforts to regain the goodwill of millions as public dissatisfaction became clear.

Planning for a crisis is not evidence of one

It is hard to judge how realistic the scenarios are. My read is that the exercises say more about what worries AI executives than about the probability of a catastrophe or uprising. Planning for a crisis is sensible; the more revealing question is what these companies count as a credible threat, and how they distinguish public protest from danger.

The reported monitoring system and the new exercises point to two different ways of preparing for unrest: tracking it in advance, and rehearsing a response if it escalates. That combination makes the companies’ posture worth scrutinizing, even without evidence that the worst-case scenarios are likely.

Elsewhere, protesters disrupted a party in support of a data center, and security escorted them out.